AI Paper Atlas

Research map / Large language models / Agents, code and math

LogicSAGE: Neuro-Symbolic Reasoning with Socratic-Guided Enhancement

Jinlong Tian, Jiang Yu, Kewei Cheng, Fengxiang Cheng, Yue He, Yunfei Wang, Haotian Wang, Haoxuan Li, Wenjing Yang, Shixuan Liu

ICML 2026 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add reactive error-guided correction (ablation)LogicSAGE (SFT Qwen3-8B) · 5 logic benchmarks avgexecution accuracy: 70.1 -> 88.4Improved
add reactive correction (ablation)LogicSAGE (SFT Qwen3-8B) · AR-LSATexecution rate: +63.6Improved
add proactive uncertainty probing on top of reactive (ablation)LogicSAGE (SFT Qwen3-8B) · 5 logic benchmarks avgexecution accuracy: 88.4 -> 91.2Improved
add proactive probing (ablation)LogicSAGE (SFT Qwen3-8B) · AR-LSATexecution accuracy: 66.5 -> 74.6Improved
add proactive probing (ablation)LogicSAGE (SFT Qwen3-8B) · PrOntoQA, ProofWriter execution rateexecution rate: +0.0No clear effect
add symbolic System 2 to neural System 1 (fusion) (ablation)LogicSAGE · 5 logic benchmarks avgaccuracy: 80.18 -> 92.36Improved
add neural fallback and arbitration to symbolic solver (ablation)LogicSAGE · 5 logic benchmarks avgaccuracy: 90.54 -> 92.36Improved
add neural fallback and arbitration (ablation)LogicSAGE · AR-LSATaccuracy: 71.32 -> 77.00Improved
supervised fine-tuning on solver-verified dataLogicSAGE (Qwen3-8B) · 5 logic benchmarks avgaverage accuracy: 80.30 -> 92.36Improved
increase refinement depth Tmax from 0 to 3 (ablation)LogicSAGE · 5 logic benchmarks avgaverage accuracy: 79.30% -> 92.36%Improved
increase refinement depth Tmax from 3 to 4 (ablation)LogicSAGE · 5 logic benchmarks avgaverage accuracy: < 0.2%No clear effect
swap backbone from Qwen3-8B (training-free) to GPT-4LogicSAGE · 5 logic benchmarks avgaverage accuracy: 80.30 -> 86.10Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy